Skip to main content
After a run, you can keep talking to any simulated participant who took part. Ask them about a moment, the exact wording they read, or why they gave up, and they answer in their own voice from what happened in their session. Follow-up interviews live in the ish web app, on the Chat tab of a participant. There is no CLI command or MCP tool for them, and API keys cannot reach them.
This is a conversation with one participant about a run that already happened. To put a new question to many people at once, use an ask. To ask every participant fixed questions at the end of a run, add interview questions to the study.

​
Before you start

  • A study with at least one simulated participant whose run has ended: completed, failed, or cancelled. Participants still running cannot be interviewed yet.
  • You are signed in to the web app. Share links are read-only and do not show the Chat tab.
Real (human) participants, participants in an ask, and participants whose person was deleted cannot be interviewed.

​
Open the chat

1

Open the study's results

Any results view works: the participants table, the overview, journeys, the timeline, frames, or segments.
2

Select a participant

Their details open in a panel.
3

Open the Chat tab

The first time, it reads “Talk to the participant” and offers two starter questions for the study’s modality. Tap one to send it, or type your own.
Press Enter to send and Shift+Enter for a new line. Replies stream in; Stop ends a reply early. If a reply fails or you stop it, Try again asks the same question again. Each message can be up to 8,000 characters of text; attachments are not supported.

​
Ask questions that use what they remember

The participant answers from a record of their session: who they were at the time of the run, what they did step by step, what they noted to themselves, their answers to the study’s questions, and their own end-of-session summary. A failed or cancelled run is part of that record, so a participant who barely started says so rather than inventing a session.
  • Point at a moment. “Take me to the moment you first hesitated” gets a more useful answer than “How was it?”
  • Ask why. “What exactly made you lose confidence?” draws out the reasoning behind a reaction.
  • Ask for the exact detail. For a price, a label, or a line of copy, the participant looks back at the screen, image, page, or clip they saw. If it was too small or blurry to read, they say so instead of guessing.
  • Hypotheticals are guesses. Ask about something outside the session (a change you are considering, new copy you paste in) and they answer as themselves but say it is a guess, not a memory. Asked about something they never ran into, they say they do not recall it.
Each reply sees the most recent 40 messages of the conversation, so in a long thread the earliest turns stop informing the answers.

​
Work on one thread as a team

Each participant has one thread, shared by everyone who can open the study. Reopen the tab and the conversation picks up where it left off. When your organization has more than one member, each question shows who asked it. Only one reply is generated at a time per participant. There is no way to clear or export a thread.

​
Cost and limits

Answers are generated from the recorded session; they are a reconstruction, not a replay. The participant’s record is fixed at the run, so they know nothing of later iterations, and the conversation does not change the run’s recorded journey. Read their answers with the same care as the rest of the results.

Reactions and results

The recorded journey a participant answers from.

Assignment and questionnaire

Questions every participant answers at the end of a run.

Ask and round

Put a new question to a whole panel instead of one person.